I can't thank the experts enough for their clarifications. One final question following my earlier posts on forming coomposite variables here, here, and here.
If I have measured some variables as cardinal numbers (e.g. 1, 2, 3, 4, 5 times) and others on a Likert scale, can I then just add the scores together?
For example:
- How many times do you visit the local hospital per month? Answer: 5
times - How would you rate the medical service provided to you: Answer
3 on a scale of five going from poor to excellent.
Can I add 5 and 3 and say the composite score is 8?
What if the likert scale goes from negative to positive (e.g. strongly disagree = -2, disagree = -1, neither disagree or agree = 0, agree = 1 and strong agree = 2)?
Would the same rule as above apply if the score is say 0 or -2?
My objective is to keep the aggregation process simple and not get 'confused' in complex statistical formulas.
Best Answer
Combining likert items with different numeric scalings
Combining count variables with likert items
log(counts + 1)
) or some similar transformation that reduces the positive skew of the distribution.log(count + 1)
or something similar instead of the raw count in your factor analysis.Potential issues with combining items on different scales